Early Discovery of Disaster Events from Sensor Data Using Fog Computing

Abstract

Disaster event such as hurricane, blizzard, and winter storm always demand an early response. The lesser the time it takes to respond, the more damage can be prevented. In a disaster event, predicting the happening and alerting the concerned authorities should be done with a minimal latency. Today’s existing technologies highly rely on information disposal to a far away control station. Hence, we aim at achieving an almost zero latency in Natural Disaster discovery. In this paper, the early discovery of disaster events are achieved with the help of Fog Computing infrastructure. Here, we have proposed a machine learning based prediction with Weather sensors. Machine Learning as a tool offers quick and highly reliable predictive models. These models once trained, it will make use of the basic computational operations. Hence they are perfectly suitable for various emergency situations. With Fog computing, the latency in data upload has been minimized. Also, for prediction purpose we have used data from over 5 years by using a Weather API. Multiple machine learning models were trained on this data, and the best model in terms of computation time has been deployed for evaluation. Our evaluation metrics show an impressive 96% accuracy of the deployed model and the response time remains as less as milliseconds.